Abnormal Behavior Detection For Patients In Nursing Rehabilitation Center

Xin Wang ； Xiaoyan Zheng ； Jianshun Liu ； Baolong Yuan ； Languang Zhao ； Jianping Sun

《淡江理工學刊》 26卷7期 (2023/07) Pp. 923-931

https://doi.org/10.6180/jase.202307_26(7).0003

摘要

This paper proposes an abnormal behavior detection using visual monitoring in rehabilitation center. The proposed method solves the problems of inaccurate and delayed detection of abnormal behavior in traditional algorithms, and provides necessary technical support for medical staff to quickly deal with patient abnormal behaviors. Firstly, the deep neural network model is adopted to detect the patient feature nodes and skeleton in the images. Then, the vector representing the position and direction of the patient limb in each frame is extracted. Finally, the angle between the trunk, legs and the ground during the fall, the aspect ratio of the body frame and the irregular motion of the wrist and ankle joins are used as discriminant features to determine the occurrence of patient fall and twitch. The experimental results show that the detection accuracy of the proposed method is 96%, and the detection speed in the real medical rehabilitation center is 25 f/s. The proposed method can monitor the patient behavior characteristics in the elderly rehabilitation center in real time, and issue an alarm timely in the case of accidental falls and twitch. Meanwhile, this paper provides a more accurate and convenient computer-aided nursing method for medical staff to monitor patient abnormal behaviors.

關鍵字

Abnormal Behavior Detection ； Convolutional Neural Network ； Medical Care ； Visual Image Processing